Abstract

BackgroundGiven the struggle in the field of psychiatry to realize the precise diagnosis and treatment, it is in an urgent need to redefine psychiatric disorders based on objective biomarkers. The results generated from large psychiatric genomic consortia show us some new vantage points to understand the pathophysiology of psychiatric disorders. Nevertheless, how to relate these captured signals to the more refined disease dimensions has yet to be established.\n\nMethodsWe chose a top-down, cross-disorder approach by using the summary statistics of GWAS from large psychiatric genomic consortia to build a genomic structural equation model (SEM) for SCZ, BD and MDD to detect their common factor (CF), and to map a potential causal core gene for the CF, followed by the transcriptional prediction of the identified causal gene in our sample and the discovery of new biotypes based on the prediction pattern of the causal gene in the brain. We then characterized the biotypes in the context of their demographic features, cognitive functions and neuroimaging traits.\n\nOutcomesA common factor emerged from a well-fitting genomic SEM of SCZ, BD and MDD (loading 0.42, 0.35 and 0.09 for SCZ, BD and MDD, respectively). One genomic region in chromosome 6 was implicated in the genetic make-up of the common factor, with fine-mapping analysis marking ZNF391 as a potential causal core gene (posterior possibility = 0.96). Gene expression inference analysis identified eight brain regions showing different expression levels of ZNF391 between patients and controls, with three biotypes arising from clustering patients based on their expression pattern of ZNF391 in the brain. The three biotypes performed significantly differently in working memory (PDMS_TC = 0.015, PDMS_TC_A = 0.0318, PDMS_t0D = 0.015) and demonstrated different gray matter volumes in right inferior frontal orbital gyrus (RIFOG) in the same order as working memory (biotype 3 > biotype 2 > biotype 1, PRIFOG = 0.0027). Using ZNF391 as instrumental variable (IV), a partial casual path could be linked from RIFOG to working memory ({beta}RIFOG->DMS_TC0D = 4.95, P = 0.0056; {beta} RIFOG->DMS_TC = 2.53, P = 0.059; {beta}RIFOG->DMS_TC_A = 2.57, P = 0.056).\n\nInterpretationThe general predisposition to several psychiatric disorders may be influenced by variations of ZNF391, through its effects on right inferior frontal orbital gyrus and working memory. This illustrates the potential of a trans-diagnostic, top-down approach in understanding the commonality of psychiatric disorders.\n\nEvidence before this studyThe results from recent cross-disorder genome-wide association studies (GWAS)using large samples indicate that there is notable genetic overlapping between psychiatric disorders. However, the structural relationship of these disorders at the genomic level and the details of refined disease dimensions affected by the associated loci in a cross-disorder pattern remains unknown. We searched the published studies (up to Sep 7, 2019) in PubMed using the combination of the following keywords \"((cross disorder) OR (schizophrenia AND bipolar disorder AND major depressive disorder) AND (genome AND structural equation) AND (cognition OR imaging))\", no published study was found. We then removed the term \"structural equation\", 23 original studies were found. To the best of our knowledge, none of these studies explored the organized structure between three disorders. Further, of 23 articles we found, the majority of them took an approach of either polygenic risk score (PRS) or candidate gene to test the association with either psychological traits such as loneliness or neuroimaging measures in one (schizophrenia) or two (schizophrenia and bipolar) disorders. Hitherto, no study has been conducted to redefine three disorders based on the biological markers generated from the cross-disorder genomic studies.\n\nAdded value of this studyAdopting a novel approach of genomic structural equation modelling, we used the latest results of GWAS of three major psychiatric disorders to detect their common factor, further, to identify the loci associated with such as a common factor, and the locis transcription consequences in the brain. Propelled by the phenomenon \"genes do not read DSM\", we used a cutting-edge clustering algorithm to redefine three disorders based on the cerebral spatial expression pattern of associated core gene. Our study provides another piece of evidence as to the potentials of utilizing the signals arising from large population-scale GWAS to dissect and redefine psychiatric disorders.\n\nImplications of all the available evidenceConsistent with previous case-control cross-disorder GWAS, our study suggests that a common factor exists in three major psychiatric disorders and the biological information of core gene associated with the common factor could be used as an objective marker to explain three disorders and their pathophysiology.